Questions
•What is learned?
–Prototype?
•Not just prototype.
•Examples plus prototype.
–Distribution?
•Gaussian?
•Seems like examples mattered.  Could distribution overfit the data?
–Examples + Weighted distances.